Steve "Guggs" Guggenheimer's Blog https://docs.microsoft.com/archive/blogs/stevengu/feed.xml Thu, 20 Jun 2019 22:41:43 GMT en-US hourly 1 Microsoft’s MT-DNN Achieves Human Performance Estimate on General Language Understanding Evaluation (GLUE) Benchmark https://docs.microsoft.com/archive/blogs/stevengu/microsoft-achieves-human-performance-estimate-on-glue-benchmark Thu, 20 Jun 2019 15:22:13 GMT https://blogs.msdn.microsoft.com/stevengu/?p=1925 GLUE) is a well-known benchmark consisting of nine NLU tasks, including question answering, sentiment analysis, text similarity and textual entailment; it is considered well-designed for evaluating the generalization and robustness of NLU models.  Since its release in early 2018, many (previous) state-of-the-art NLU models, including BERT, GPT, Stanford Snorkel, and MT-DNN, have been benchmarked on it, as shown in GLUE leaderboard. Top research teams in the world are collaboratively developing new models, approaching human performance on GLUE. In the last few months, Microsoft has significantly improved the MT-DNN approach to NLU, and finally surpassed the estimate for human performance on the overall average score on GLUE (87.6 vs. 87.1) on June 6, 2019. The MT-DNN result is also substantially better than the second-best method (86.3) on the leaderboard.
The snapshot of the GLUE leaderboard on June 6, 2019
The latest improvement is primarily due to incorporating into MT-DNN a new method developed for the Winograd Natural Language Interface (WNLI) task in which an AI model must correctly identify the antecedent of an ambiguous pronoun in a sentence. For example, the task provides the sentence: “The city councilmen refused the demonstrators a permit because they [feared/advocated] violence.” If the word “feared” is selected, then “they” refers to the city council. If “advocated” is selected, then “they” presumably refers to the demonstrators. Such tasks are considered intuitive for people due to their world knowledge, but difficult for machines. This task has proven the most challenging in GLUE, where previous state-of-the-art ML models can hardly outperform the naïve baseline of majority voting (scored at 65.1), including BERT. Although the early versions of MT-DNN (documented in Liu et al. 2019a and Liu et al. 2019b) already achieve better scores than humans on several tasks including MRPC, QQP and QNI, they perform much worse than humans on WNLI (65.1 vs. 95.9). Thus, it is widely believed that improving the test score on WNLI is critical to reach human performance on the overall average score on GLUE. The Microsoft team approached WNLI by a new method based on a novel deep learning model that frames the pronoun-resolution problem as computing the semantic similarity between the pronoun and its antecedent candidates. The approach lifts the final test score to 89.0.  Combined with other improvements, the overall average score is lifted to 87.6, which surpasses the conservative estimate for human performance on GLUE, marking a milestone toward the goal of understanding natural language.
The Microsoft team (from left to right): Weizhu Chen and Pengcheng He of Microsoft Dynamics 365 AI, and Xiaodong Liu and Jianfeng Gao of Microsoft Research AI.
Previous blogs on MT-DNN could be found at: MT-DNN-Blog and MT-DNN-KD-Blog. Cheers, Guggs  ]]> Business Applications Sessions at Build 2019 https://docs.microsoft.com/archive/blogs/stevengu/bizapps-sessions-at-build-2019 Thu, 02 May 2019 08:30:32 GMT https://blogs.msdn.microsoft.com/stevengu/?p=1915 th when Satya starts the day with the Vision Keynote I will be just as excited as I have been in previous years. Build provides an amazing stage to highlight the incredibly broad range of building blocks that developers are leveraging, including what’s new in the area I’m currently focused on which is the Business Applications and Power Platform space. For those of you who will be attending Build and are interested in diving deeper into these areas, here are a few sessions that you might be interested in, including an end to end overview of this area that I’ll be leading on Monday afternoon. Future Business Apps in Power Platform, Dynamics and Office Monday May 6, 2019 2:00pm – 3:00pm https://mybuild.techcommunity.microsoft.com/sessions/77082 I will be leading a session with Orville McDonald and Justyna Lucznik to discuss how to create the latest in business applications. It will be an end-to-end session starting with getting insights from your data using AI and Power BI. Then it is followed up with using the Power Platform to connect data, create apps with no “cliff development” and then get deeper into the code to create a custom app. AI in Power BI Wednesday May 8, 2019 10:00am – 11:00am https://mybuild.techcommunity.microsoft.com/sessions/77088 For the past few years I have seen many companies create AI solutions without first getting their data estate in order or getting the insights to determine how best to prioritize their scenarios. Power BI is a great way to do BI before AI and now with AI infused experiences in Power BI you can get deeper insights into your data. Building an App on the Power Platform Wednesday May 8, 2019 3:30pm – 4:30pm https://mybuild.techcommunity.microsoft.com/sessions/77091 The Power Platform can help you accelerate the time it takes to get your app to market with a development platform that enables you to do a lot with no-code/low-code tooling. In this session learn how to develop business apps, extend with custom visuals and then to package and publish on AppSource for distribution. PowerApps Component Framework Wednesday May 8, 2019 9:30am – 9:50am https://mybuild.techcommunity.microsoft.com/sessions/77350 In this session you can learn more about the PowerApps component framework including developer tooling, how to build your first control and some examples of what others have built. Embed Low-Code/No-Code Capabilities in your App with PowerApps and Flow Tuesday May 7, 2019 8:30am – 9:30am https://mybuild.techcommunity.microsoft.com/sessions/77090 Developing Business Applications via Microsoft Power Platform Tuesday May 7, 2019 3:30pm – 4:30pm https://mybuild.techcommunity.microsoft.com/sessions/77155 Building Azure Apps using the Common Data Service Tuesday May 7, 2019 5:00pm – 6:00pm https://mybuild.techcommunity.microsoft.com/sessions/77085 Building Data Connectors for the Microsoft Power Platform Tuesday May 7, 2019 10:00am – 11:00am https://mybuild.techcommunity.microsoft.com/sessions/77092 Dataflows and Data Interop with the Common Data Model Wednesday May 8, 2019 8:30am – 9:30am https://mybuild.techcommunity.microsoft.com/sessions/77087 Intelligent Process Automation with Microsoft Flow Wednesday May 8, 2019 12:30pm – 1:30pm https://mybuild.techcommunity.microsoft.com/sessions/77244 Sprinkle some DevOps on your PowerApps & Dynamics 365 Customer Engagement projects Wednesday May 8, 2019 5:00pm – 6:00pm https://mybuild.techcommunity.microsoft.com/sessions/77093   You can learn more about these Build sessions and find more in the session catalog. I hope that you enjoy your time at Build 2019, and we look forward to seeing what you create. Cheers, Guggs]]> New Program for Business Applications ISVs https://docs.microsoft.com/archive/blogs/stevengu/new-program-for-business-applications-isvs Wed, 17 Apr 2019 12:38:13 GMT https://blogs.msdn.microsoft.com/stevengu/?p=1905 Note that this blog post was original posted to: https://www.linkedin.com/pulse/new-program-business-applications-isvs-steven-guggenheimer   The growth of data and the evolution of software as a service (SaaS) applications is driving digital feedback loops that are changing how people do their work and drive their businesses. We’ve been through many iterations of both our infrastructure and our approach to building and delivering solutions. The journey for Office from client, to client-server, to SaaS has been well documented as has the evolution from our Windows Server family of offerings to Azure. With the April ’19 release of Power Platform and the updated Dynamics 365 offerings, the business application family is now well into its transformation from client & client-server offerings to a modern SaaS platform (Power Platform) and modern SaaS service (Dynamics 365). This transformation is also well documented. As we make the Platform and SaaS transformation in this space, the time is right to modernize our independent software vendor (ISV) approach and program to support ISVs in our joint effort to accelerate deployment of PowerApps and Dynamics 365-based solutions. From how we provide better tools in the platform for ISVs and developers, to new program offerings, and alignment with Microsoft’s broader ISV efforts, we are looking holistically at how we can better serve our ISV partners. In talking with ISVs over the last few months there is excitement for the future while acknowledging that changes are needed to achieve it. As a platform company we know our success is determined by the health of the ecosystem and as such we wanted to give you a preview of the work we are doing that will start rolling out in July. You can learn more about the program at: https://cloudblogs.microsoft.com/dynamics365/bdm/2019/04/17/upcoming-program-for-business-applications-isvs/ Cheers, Guggs]]> Accelerating the ISV Opportunity https://docs.microsoft.com/archive/blogs/stevengu/accelerating-the-isv-opportunity Tue, 05 Feb 2019 09:19:11 GMT https://blogs.msdn.microsoft.com/stevengu/?p=1895 hardware (OEM) and software (ISV) ecosystems, I know first-hand how hard we work to make partners successful leveraging Windows, Azure, and the other Microsoft products and how this work continues today through both our OEM and One Commercial Partner (OCP) organizations. As we look forward, we want to build on the work we’ve done over the last 5+ years working with the software as a service (SaaS) and line of business (LOB) ISV ecosystem around Azure and Office 365 to extend this work to do more with Dynamics 365 and the Power platform. As we transition the older Dynamics offerings from a series of monolithic applications to a set of SaaS offerings built on a uniform application platform, it presents a new opportunity for us to partner with the ISV community both on the product/platform capabilities and on our go to market (GTM) efforts. First some background. Office, Dynamics and Windows went through multi-year journeys to transition from on-premises offerings to SaaS offerings. James Phillips recently wrote about the digital feedback loop from the Power platform and how that has empowered people to change their organization with software that would not have previously been possible for them. The insights derived from Power BI are made actionable by applications created by PowerApps and then Flow simplifies robotic process automation (RPA) driving feedback to Power BI and continuing the digital feedback loop. At the core of these platform assets is data. To maintain a consistent data model, the Power platform supports the common data model (CDM) where you can leverage or define a consistent data scheme for entities. By defining it early you can then create producers and consumers of data and have them work together on the same data set making the whole more valuable than the individual applications. For LOB ISVs this creates an opportunity for building on top of this data asset, connecting additional data sources or embedding it in other applications. Furthermore, Microsoft products historically had their own extensibility model so that someone wanting to extend Office would have to work with a different extensibility model than someone extending Dynamics 365. Satya Nadella recently talked to the press about how the Power platform is the extensibility model for Microsoft 365 and Dynamics 365 along with being able to integrate with 3rd party tools. Looking through the April 2019 release notes shows what we are working on for our next release. I recommend that ISVs look at the notes to see the upcoming updates to identify new opportunities for their business. As an example of how ISVs are starting to leverage the Power platform and Dynamics 365 we can look at  Indegene a partner that develops solutions for healthcare and pharmaceutical enterprises globally. They combined Dynamics 365, Power platform and Azure to bring a new customer engagement solution to the life sciences/healthcare industry by extending the CDM for life sciences sales, marketing and account management; added numerous commonly used business processes including a mobility component; and embedded descriptive as well as prescriptive intelligence capabilities that a life sciences manufacturer would typically not have as a part of their core solution. Taking advantage of the broader Azure platform they have included services for AI (natural language understanding, speech, graph database and more) which work with the data model and business logic in Dynamics 365 and the Power platform. With minimal development Indegene was able to create intelligent assistants and customer level personalization in life sciences. Without these Microsoft capabilities, this type of enterprise solution would have taken Indegene many years to develop. As Sanjay Virmani, EVP at Indegene recently said, “as customer engagement with healthcare providers and patients is moving towards becoming more intelligent than merely transactional, we saw a significant opportunity to disrupt the space and having a full toolbox between Dynamics 365 + Power platform + Azure has been the key to our substantive progress.” We recognize that technology is only part of what makes ISVs successful, it is important that the business side is equally as robust. In support of this we are complementing the ongoing engineering work with an additional focus on alignment with our OCP, Azure and field teams to create the right GTM motions necessary for ISVs. Changes in the underlying technology often leads to taking a first principles approach to creating new software and we are doing the same exercise for our tools and resources for ISVs. I think of this in terms of key focus areas that include: Developer Tools - The revamping of our self-service tools for developers to learn about and leverage the Dynamics 365 solutions and the Power platform. We have a lot of work to do here in conjunction with our April release cycle, but know we are focusing on it. This also includes our work on accelerators, development centers, certification and all the other support tools/services we can provided ISVs. AppSource - We will align all our marketplace efforts with the work the Azure team is doing on Azure Marketplace and our joint work on a modern commercial marketplace for the benefit of both our AppSource customers and for ISVs. This includes both the technical and marketing benefits for both AppSource and Azure Marketplace. Being able to publish once to merchandize across storefronts to all Microsoft’s customers, sellers and partners will open new growth opportunities to most ISVs. GTM Alignment – We have the opportunity to work with the OCP team on the GTM work we’ve been doing with Azure across both marketing tools and co-sell motions. Aligning our efforts in this area will bring the Dynamics 365/LOB ecosystem support in line with our Azure GTM efforts. New Offerings – Along with alignment with the company’s current GTM efforts, are there things we can learn from other companies in the industry and bring them forward to our partners. Given all the changes we are working on, we are using this time to look at what we can learn from others that could be applied to our ecosystem. So, as we look toward to the April release and our new fiscal year, know that the teams are focused on ISV support for the evolution of Dynamics 365 and the Power platform, alignment with the Azure AppSource efforts, coordinating with other Microsoft GTM motions and considering new best practices. We are taking feedback from ISVs to make sure that the marketplace has the benefits and GTM strategies that they need to maximize success in addition to looking at the platform to make sure there is enough surface area for them to be successful. You can hear some more of my thoughts on supporting ISVs as part of a broader conversation on the Steve Mordue podcast. In February I will be attending Business Forward in Paris and Mobile World Congress in Barcelona if you would like to meet about your ISV. Cheers, Guggs]]> LOB ISV Engagement for Dynamics 365 and the Power Platform https://docs.microsoft.com/archive/blogs/stevengu/isv-for-dynamics-365-and-the-power-platform Tue, 08 Jan 2019 10:03:48 GMT https://blogs.msdn.microsoft.com/stevengu/?p=1885 Note: this post was originally posted here. Throughout my career I have had the opportunity to work with businesses, partners and public entities around the world to leverage Microsoft technologies within their organizations. I’ve also worked closely with our ecosystem partners; both hardware (OEM) and software (ISV), to team up with Microsoft in the creation of their offerings and the growth of their businesses. These engagements have spanned from two-person start-ups to Fortune 1000 corporations and everyone in between. Working with the ecosystem to drive mutual success and watching people and companies grow has been one of my favorite areas to focus on over the last 20+ years at Microsoft. In that spirit, I’m happy to add working with ISVs back into my remit as a complement to the work I’ve been doing around AI. Starting in November 2018 my role expanded to include engagement with the broad set of Line-of-Business (LOB) ISVs that build on the Microsoft platforms, with a focus on growing the Dynamics 365 ISV community. This continues to be an incredible time for ISVs as both customers and partners continue their digital transformation and look to their ISVs to help them along the way. The latest step in that journey for Microsoft has been the transformation of the Dynamics 365 product line to a set of business SaaS offerings along with the creation of a platform asset (Power platform) that all LOB solutions can leverage in their transformation. Given our experience in transforming our own platforms and solutions from on-premise offerings to cloud offerings, we look forward to continuing helping ISVs transform and grow based on the state of their solutions and customer base. In this initial post I would like to share my perspective on how the team and I are looking at this opportunity and our plan to help existing and new partners take advantage of what we’ve learned to date. Over time we’ll provide updates on the tools available to help ISV’s from both a technical and go-to-market perspective. Note this new series of blog posts will complement the ongoing AI blog series, and not replace it.

Opportunity

As we work with customers and partners it is clear the opportunity for LOB ISVs continues to grow.  Whether it’s by industry or horizontal solutions, migrating customers from existing on-premise solutions to the cloud or expanding a customer’s cloud portfolio, or helping customers leverage the new generation of building blocks, the opportunity for ISVs has never been greater.  At the same time, there are many ISVs that are transforming their business by leveraging new technologies and looking to expand their footprint. In all these cases we want to add the Dynamics 365 and Power platforms into the suite of tools that Microsoft can offer to help ISVs in their efforts. Microsoft at its core is a platform company and the transition of the Dynamics 365 family into a SaaS platform and a set of SaaS services that can be customized nicely complements Azure and Office to provide the only integrated set of platform offerings from IaaS to PaaS to SaaS that ISVs can leverage in their work. The focus of our efforts in this area will be to ensure the LOB ecosystem is able to leverage any aspect of the platform that might be useful. Whether that is the Common Data Model and Common Data Service, the Power platform including Flow, PowerApps and Power BI, or the Dynamics 365 applications (sales, customer service, finance and operations, etc.). There are even opportunities to leverage the new generation of building blocks including AI and Mixed Reality. As an organization focusing on ISVs, we have three core focus areas: 1) find and partner with ISVs that will help our mutual customers be successful, 2) minimize the friction to onboarding to Dynamics 365, Power platform and AppSource; and 3) help ISVs that build on the platform grow their business. We believe there are clear benefits for ISVs who work with us in this area that include:
  • Expanding your Business by leveraging the Microsoft Partner ecosystem and our best in class sales and go-to-market support.
    1. Promote your app on AppSource to reach up to 100 million commercial active users.
    2. Join a global community that connects you to the relationships, insights, tools, resources, and programs to expand customer reach.
    3. Leverage Microsoft co-selling programs and resources to drive new revenue growth.
  • Building Modern Cloud LOB Solutions on Microsoft’s Azure, Office 365 and Dynamics 365 offerings, and leverage the Microsoft Power platform and new technology building blocks.
    1. Innovate with a common data platform that brings together your customers’ data with unique Microsoft data assets.
    2. Enable app development from citizen to professional developers through the Power Application tools.
    3. Leverage existing solution frameworks in Marketing, Sales, Service, Commerce, Operations, Finance, Talent (MSSCOFT).
    4. Powerful new building blocks - AI, Mixed Reality, IoT.
  • Getting to Market Faster on a trusted global platform with familiar tools, unique data assets and industry accelerators.
    1. Take advantage of your team’s existing knowledge of Azure and Office 365, as well as vertical accelerators.
    2. Protect and govern your sensitive data using Azure’s industry-leading security and the most comprehensive compliance offerings.
    3. Our global infrastructure spans more than 100 highly secure facilities worldwide while meeting local data residency needs.

Using the Dynamics 365 and Power Platforms

As we work with LOB ISVs that are currently leveraging the Dynamics 365 and Power platforms there are a set of common patterns we are seeing in terms of how they leverage these offerings:
  • Build: There are ISVs leveraging the Power platform (e.g. PowerApps, Power BI, etc.) in the creation of their SaaS offerings. In this pattern ISVs build unique horizontal and/or vertical solutions that have their own user experience and standalone functionality.
  • Embed: There is another set of ISVs we see building on the Dynamics 365 first-party apps (e.g. Dynamics 365 for Sales) to create industry focused apps. They have a high-level dependency on Dynamics 365 and leverages the same UI model as the app they build on. Often, they also build on the Power platform to create additional functionality/capability.
  • Connect: The third pattern we see is when ISVs build stand-alone experiences that connect into Dynamics 365 or the Power platform. These connected offerings enhance the Dynamics 365 capabilities by providing additional capabilities that are not natively part of the platform. These offerings have limited dependency on the platform and have their own user experience.

ISV Success

At the end of the day, any platform is judged by the health of its ecosystem. With Dynamics 365 and the Power platform providing a rich set of capabilities for ISVs to leverage, we are focused on making these tools a successful part of our ecosystem offerings. The team and I will be the champion for LOB ISVs and developers in this space, representing their unique needs inside of Microsoft and in the broader community. I look at this extended remit as an opportunity to knit together the work of our ISV focused teams, programs and tools with a goal of creating a unified frictionless experience for our partners. By working to ensure ISVs can quickly build quality applications, with the right knowledge and guidance, as well as have access to GTM programs that enable ISVs to grow their overall business we want to be a great partner for the ISV community. I will leverage my experience from OEM and DX while focusing on unique opportunities for Dynamics 365. This week I will also be attending CES 2019 through Wednesday in Las Vegas talking with companies about AI, but also available to any LOB ISVs that would like to meet. I will also be at Mobile World Congress 2019 in February as well. Cheers, Guggs  ]]> The AI Journey https://docs.microsoft.com/archive/blogs/stevengu/the-ai-journey Thu, 01 Nov 2018 11:30:25 GMT https://blogs.msdn.microsoft.com/stevengu/?p=1815 patterns for AI, but the journey itself. This notion of real world or pragmatic AI stems from conversations with many customers and partners and their efforts to date in the last year. These projects typically align well with the patterns we’ve discussed, have been undertaken to both understand what is possible with AI and to learn more about the tools and platforms that are available. In many instances teams would take a set of data offline, build a model, a tool, or a service/solution, test it out, even put it in production as a proof of concept (PoC) and drive some learning. Unfortunately, many of these projects don’t feed into the core data estate or pipeline, they didn’t quite generate enough ROI to justify scaling the work further, nor did they create a sustainable differentiation for the company.  So, while they were great for learning, which is important in itself, they were not great for creating a sustainable AI asset.  With good learning done, the question now becomes the best way to think about the where and how of creating a sustainable AI asset/capability that support the differentiation and growth of an organization. In this area we’ve created a framework to think about that AI Journey that I recently tested with customers in Europe and here in Redmond. This framework seems to resonate very well with the people we’ve spoken with, so I wanted to share it more broadly and hear what others think. Note that the journey is not necessarily linear, it’s more about picking the right tools and entry point(s) for any given organization, and scenario, and to think about what makes the most sense from an investment and opportunity cost perspective. Like any new technology the question is not how to use it in every scenario all at the same time, but rather where can it be used to help a company grow and differentiate and where do you want to build core capability versus leverage the services or tools from others. With that in mind here are the four areas we have been discussing with customers/partners.

BI before AI

Data is the foundation for AI and without data there is no AI, so the first opportunity we have is to leverage the data we have within an organization. I like to think that it’s good to look for “Insight” before we try and drive “Intelligence” and thus the notion of “BI before AI”.  Where does a company have unique data that can be leveraged?  Where are their data sources that can augment that data….is a company data driven, do they make decisions based on data, etc.  If not, that is a great place to start as data will be the cornerstone for much of a company’s AI work going forward. With the hype for AI being so strong many people want to rush into AI before making sure their data estate is well structured. This step is both critical in obvious and subtle ways. Data is used to train models so obviously you need to have enough data to train accurate models without overfitting. More subtly is that it is important to have a diversified dataset to minimize the risk of bias.  One of my colleagues Judson Althoff often says, “…without good data, all you will do with that fancy AI technology is make mistakes with greater confidence than ever before.” For larger organizations with a lot of data, or a long history of data collection, there will be work to do to determine which data to try and clean/normalize and use….and where to start over collecting data from scratch, in a way that best serves the insights and ultimately intelligence they want to drive.  So having a plan for the data estate, and a path that balances driving insights while building out a sustainable data asset is a great entry point on the AI journey. Along with cleaning and normalizing the data, getting the data in a standard format (e.g. common data model [CDM]) enables you to get insights into where best to prioritize your efforts. Normalizing your data with CDM and adopting tools like PowerBI makes it possible to not only get insights but to also easily create PowerApps and line of business tools.

SaaS AI Offerings

I have personally observed that when AI is used to help support some of the core business processes it is easier to get organizational alignment and support for continued investment, which is always critical when seeking to invest new areas like AI. In this area we are starting to see teams use AI to augment standard business processes, and within Microsoft our team incubated 3 AI solutions (customer service, sales, marketing) in the last year that are now starting to become available to others through the Dynamics 365 AI offerings. In many cases, we start by applying AI on top of our BI. That is automating the understanding of what is inside the data we are collecting. The simple example of this is the use of AI to look for patterns and signals in the new customer service insights.  On a daily basis the AI does clustering on the last time window of topics and compares that with other BI data to identify emerging trends. By using AI to assist professionals in common business processes and scenarios, the organization sees real impact from AI and provides some proof points to consider for different use cases in the company. The previous examples were role specific but you can use AI to get a broader understanding of the organization with Workplace Analytics. In addition, there are many third-party SaaS services being created to support different horizontal and industry specific processes. These will become more important over time as the opportunity to purchase SaaS services with AI built in, or available as an add-on, will likely become more economical than trying to build your own AI capability for every business area/process within your organization.

AI Accelerators

Up until this point we have normalized our data to get insights on where to invest our efforts and leveraged AI in products that make core business processes more effective or efficient. The next opportunity is to leverage the new AI layer in the development stack to develop more custom solutions or services. In this area, the ability to leverage the tools within Azure AI to augment existing processes or solutions can be very powerful and are being tested within many organizations today. There are many examples within the Azure AI Gallery to explore what is possible. The capabilities offered through the cognitive services for vision, text, speech or knowledge, the BOT service, Azure ML and more provide a great starting point for infusing AI into your core offerings. In the same way Microsoft is using the Azure AI platform to infuse AI into our own products and other companies are also able to do the same thing for their products/services. In most cases, the core set of cognitive services will get you going and there is even the option to customize them when you need something unique to your use case. The advantage of this approach is that you don’t need to be a deep learning expert or have a lot of training data to get started. Simply call the API that you need or plug into the Azure Bot Service to get up and running quickly. The most common AI apps and services developed with these tools follow the 5 patterns for AI solutions.

Custom AI

Even with all the AI building blocks that make it easy to get started using AI in your apps and solutions there will be times that you may want to do something completely custom and build your own models and tools from scratch.  In those cases, we offer the data infrastructure, management tools, computational power, and frameworks to accomplish what you need. Starting with the data we support a variety of data solutions from cloud hosted databases to data lakes. Furthermore, if you started with your data in a CDM schema you can use it in Azure Data Lake Storage generation 2 providing a clean way to scale up your data capabilities as needed. When dealing with big data the ability to process it quickly is important and Azure has support for CPUs, GPUs and FPGAs. For those of you who are not familiar with FPGAs they are customized to execute AI tasks very quickly so that you can lower your model training time but compared to CPUs they are less flexible. Today several frameworks are used for deep learning and Azure provides support for many of them. PyTorch, Keras and TensorFlow are popular ones but we also support the ONNX standard that will give your models portability if you need to move between ONNX supported frameworks.  So, the tools for creating your own customer AI solutions are woven through the Azure platform and available for developers and data scientists to take advantage of today.

Summary

With the continuing evolution of AI, the opportunity to begin applying it to real world problems is here.  Since there are multiple entry points for applying AI, the key is to determine an approach that creates both short term results and builds a long-term asset. Part of achieving that goal is determining where and how to leverage AI, and like all other investment decisions where to apply your scarce resources and where to leverage other tools or services to drive business outcomes. It’s also critical to think about where you have unique data assets and how to bring those into play as part of your overall AI journey and strategy. As you consider your next AI project start with the business outcome you are trying to drive, the depth of AI experience and the level of AI customization needed will help you determine where best to start.  As I often say, “it is too early to do everything with AI but too late to do nothing” so get started and we look forward to seeing what you develop. Cheers, Guggs (@StevenGuggs)]]> Microsoft Tops Four AI Leaderboards Simultaneously https://docs.microsoft.com/archive/blogs/stevengu/microsoft-tops-ai-leaderboards Wed, 26 Sep 2018 09:00:08 GMT https://blogs.msdn.microsoft.com/stevengu/?p=1765 Businesses are continuously trying to be more productive, while also better serving their customers and partners, and AI is one of the new technology areas they are looking at to help with this work. My colleague Alysa Taylor recently shared some of the work we are doing to help in this area with  our upcoming Dynamics 365 + AI offerings, a new class of business applications that will deliver AI-powered insights out of the box. These solutions help customers to make the transition from business intelligence (BI) to artificial intelligence (AI) to help address increasingly complex scenarios and derive actionable insights. Many of these new capabilities that will be shipping this October, are powered by breakthrough innovations from our world-class AI research & development teams who contribute to our products (including Bing and Microsoft Search) and participate in the broader research community by publishing their results. This time last year, I wrote about the Stanford Question Answer Dataset (SQuAD) test for machine reading comprehension (MRC). Since writing that post, Microsoft reached another major milestone, creating a system that could read a document and answer as well as a human in the SQuAD 1.1 test. Although this is a test which is different than real world usage we find that the research innovations make our products better for our customers. Today I’d like to share an update on the next wave of innovation in natural language understanding and machine reading comprehension. Microsoft’s AI research and business engineering teams have now taken the top positions in three important industry competitions hosted by Salesforce, the Allen Institute for AI, and Stanford University. Even though the top spots on the leaderboards continuously change I would like to highlight some of our recent progress. Microsoft tops Salesforce WikiSQL Challenge Data stored in relational databases is the “fuel” that sales and marketing professionals tap to inform daily decisions. However, understanding how to get value from the data often requires deeply understanding the structure of the data. An easier approach is to use a natural language interface to query the data. Salesforce published a large crowd-sourced dataset based on Wikipedia, called WikiSQL, for developing and testing such interfaces. Over the last year many research teams have been developing techniques using this dataset and Salesforce has maintained a leaderboard for this purpose. Earlier this month, Microsoft took the top position on Salesforce’s leaderboard with a new approach called IncSQL. The significant improvement (from 81.4% to 87.1%) in test execution is due to a fundamentally novel incremental parsing approach combined with the idea of execution guided decoding detailed in the linked academic articles above. This work is the result of collaboration between scientists in Microsoft Research and in the Business Application Group. WikiSQL Leaderboard Microsoft tops Allen Institute for AI’s Reasoning Challenge (ARC) The ARC question answering challenge, provides a dataset of 7,787 grade-school level, multiple-choice open domain questions designed to test approaches in question answering. Open domain is a more challenging approach for text understanding since the answer is not explicitly present. Models must first retrieve related evidence from large corpora before selecting the choice. This is a more realistic setting for a general-purpose application in this space. The top approach, essential term aware – retriever reader (ET-RR) was developed jointly by our Dynamics 365 + AI research team working with interns from the University of San Diego. The #3 position on the leaderboard is a separate research team comprised of Sun Yat-Sen University researchers and Microsoft Research Asia. Both results serve as a great reminder of the value of collaboration between academia and industry to solve real-world problems. AI2 Reasoning Challenge Microsoft tops new Stanford SQuAD 2.0 Reading Comprehension In June 2018, SQuAD version 2.0 was released to “encourage the development of reading comprehension systems that know what they don't know.”  Microsoft currently occupies the #1 position on SQuAD 2.0 and three of the top five rankings overall on, while simultaneously maintaining the #1 position on SQuAD 1.1. What’s exciting is how multiple positions are occupied by the Microsoft business applications group responsible for Dynamics 365 + AI demonstrating the benefits of infusing AI researchers in our engineering groups. SQuAD 2.0 Leaderboard These results show the breadth of MRC challenges our teams are researching and the rapid pace of innovation and collaboration in the industry. Combining researchers with engineering to tackle product challenges while participating in industry research challenges is shaping up to be a beneficial way to advance AI research and bring AI based solutions to customers more quickly. Cheers, Guggs]]> AI and Digital Feedback Loops https://docs.microsoft.com/archive/blogs/stevengu/ai-and-digital-feedback-loops Wed, 18 Jul 2018 08:44:33 GMT https://blogs.msdn.microsoft.com/stevengu/?p=1745 I am here in sunny…aka “hot” Las Vegs for the 2018 Microsoft Inspire and Ready conferences. Despite the heat I love this time of year for the opportunity to meet with many of Microsoft’s great partners to discuss Artificial Intelligence (AI) and the opportunities to work together in this area. Over the last year I’ve had hundreds of conversations with customers and partners on AI and what it can mean for their business/industry, which I documented in my last series of blog posts (see some of the themes here, here, and here). With the ongoing digitization of many line-of-business processes there is a growing wealth of structured and unstructured data that can be used to provide insights that improve systems and processes across core business functions.    While we have often taken an “infrastructure first” approach in the software space (e.g. client, client-service, Internet, etc.) with a new generation of technology building blocks emerging, we will see a transition to an “experiences” and “processes” first approach in the coming decade. The ability to combine core business data with a new generation of “observation-driven” data signals that will provide the starting point to drive intelligence into all of our business processes. This shift in combination with scalable cloud compute and powerful AI algorithms means organizations are better equipped than ever before to use their data to improve business outcomes, and I believe these data-powered ‘Digital Feedback Loops’ will fundamentally change how companies innovate business processes. When we couple this thinking with the AI Patterns we discussed before we can begin to see the alignment of AI with these core processes/loops. In most cases, organizations start small by creating a feedback loop around a core business process. For example, an industrial manufacturer might begin using an instrumented manufacturing line to predict breakdowns and optimize production flow. Or a bank might begin aggregating its call center activity to engage customers with more personalized offers. However, the true power of this model stretches far beyond a single feedback loop. The beauty of our growing data and AI capabilities will be the opportunity to break down silos and synthesize data across product, customer, employee, and operational business processes. In this sense, the most powerful feedback loops will be the ones that are interconnected and that reinforce each other by bringing together disparate datasets.

Customer + Product

One of the most common examples of this interconnectedness is between the processes that support customer engagement and product development. Delivering good products and intelligently engaging customers are deeply linked and complementary. Customer understanding can always help a company better tailor a product for a given market/customer, if the feedback is woven into product planning and creation. In addition, this same customer feedback can be used to deliver more effective and personalized sales, marketing, and service. When Progressive, one of the largest auto insurance companies in the USA wanted to drive deeper engagement with their customers they created Flo, a chatbot for helping customers with their questions. This is an area where we see the “Virtual Agents” pattern being one of the first implementations of AI in the customer feedback loop. Recognizing that many of their customers were using their mobile phones, Progressive used Azure Cognitive Services and Bot Framework to create the Flo Chatbot and made it available through Facebook Messenger. Now customers can get their questions answered, get quotes, and chit chat with their bot. As Progressive grows Flo’s knowledgebase over time across many customer interactions, it can use this data to improve Flo’s capabilities and even optimize the products (insurance offers) it delivers based on evolving customer needs.
”One of the great things about Bot Service is that, out of the box, we could use it to quickly put together the basic framework for our bot.“ Matt White: Marketing Manager, Personal Lines Acquisition Experience, Progressive Insurance
Developing useful bots can be challenging due to a combination of how hard it is to answer open domain questions and computers lacking conversational skills. Developing a chat bot like Flo, with its unique personality and the variety of questions that it needs to handle can be difficult, but with tools for language understanding (LUIS) and questions and answers (QnA Maker), it is easier to create innovative bot experiences quickly. Microsoft has also deeply researched social bots which provide great insights into how to build a bot that people feel comfortable interacting with. For example, Xiaoice in China with over 500M+ users is trained to have open domain conversations in a natural way, with one of the longest conversations lasting over 29 hours. Where Progressive began adopting its feedback loop with a focus on customer engagement, Tetra-Pak has taken a more product-centric approach. One of Tetra-Pak’s key business lines provides complex machinery to help food companies safely package their food. Cows typically provide 6 gallons of milk per day that needs to be packaged and delivered to customers promptly. Providing reliable products is extremely important to Tetra-Pak as any problems with the packaging process or breakdowns in the packaging line can result in large amounts of spoilage. To manage these risks, Tetra-Pak instrumented its products and started using Azure to capture the telemetry. By analyzing this data, Tetra-Pak can proactively predict product breakdowns and identify the repairs needed to prevent any issues. Not only does this help Tetra-Pak better understand its own products and opportunities to improve them, but it also allows Tetra-Pak to engage its customers more intelligently. Service engineers can be proactively dispatched and can be armed with all the right information and tools they will need to address the issue at hand. This not only drives better outcomes for the customer, but it provides them with a better experience as well.

Employees

Just as the product and customer feedback loops are linked, so too are the feedback loops companies build around their employees. For example, Tetra-Pak has taken the scenario above even further by extending its investments in product instrumentation to empower its employees. By using HoloLens, all the details on the customer and telemetry from the asset being worked on are available through the headset in real-time. Ambient Intelligence is the AI pattern where AI is applied to physical space and when telemetry is combined with this pattern it makes easier for the service engineer to help their customer. Not only does this help with lowering the time needed for a repair, but it reduces downtime and helps the service engineer gain more credibility with the customer through easy access to key product knowledge. In the most compelling scenarios, companies are empowering their employees not just with the relevant data they need, but with AI to reason over that data and assist where needed. Publicis has 80,000 employees and like other large organizations, finding the person with the right skills that you need access to can be a difficult task. To help solve this problem, Publicis teamed up with Microsoft to develop Marcel, an AI powered platform for connecting Publicis’ employees. This is an example of the pattern of “AI assisting professionals” which Publicis will accomplish in 3 ways. First, AI will be used to identify the skills and connections of each employee and then make it easier to search for and find the person that has the skills needed to accomplish a specific task. Second, by having a complete knowledge graph of the skills and abilities available within the company, collections of people that show unique combinations of expertise can be surfaced for specific projects. Third, is the ability to create new and ad-hoc working groups will allow employees can now participate in projects beyond their immediate boundaries creating more organic opportunities for employee growth.

Operations

The interconnected power of the feedback loop also extends to operations. As companies build increasingly robust data assets around their products, customers, and people, opportunities quickly appear to deliver products & services more efficiently and reliably. Think of how data + AI has changed the Taxi industry, where street-side hailing is often replaced by apps that use location data and complex algorithms to optimize the matching of riders with drivers. Jabil, one of the world’s leading design and manufacturing solution providers provides a powerful example of how the Digital Feedback Loop can also help companies optimize their operational processes. Using Azure Machine Learning, Jabil can analyze a given manufacturing process and identify high-risk areas where it will fail. By taking existing defect data and using ML techniques to create a custom “inspection model” Jabil has “seen at least an 80 percent accuracy rate in the prediction of machine processes that will slow down or fail, contributing to a scrap and rework savings of 17 percent,” said Clint Belinsky, vice president, Global Quality, Jabil. (source) Using data from sensors to optimize and maintain the health of a system is one of the ways to use the autonomous systems AI pattern. First step is to instrument your system or process to get key health metrics and then to use that data to trigger an automated system to fix problems as they arise.

Creating the Digital Feedback Loop

Although I provided a couple examples of how customers are using Microsoft AI patterns and technologies to drive feedback loops across customers, employees, products, and operations; this just the start. At this point in time it is too early for AI to solve every problem, but it is too late not to get started. Whether you are attending Inspire in person or online, take some time to attend some of the AI sessions and think about how some of these AI technologies can be used for your digital feedback loops. I am looking forward to seeing what you build. Cheers, Guggs]]> Emerging AI Patterns https://docs.microsoft.com/archive/blogs/stevengu/emerging-ai-patterns Thu, 24 May 2018 08:00:37 GMT https://blogs.msdn.microsoft.com/stevengu/?p=1715 Digital Transformation” impacts every business and every industry; from media to sports, from finance to healthcare, and from Fortune 1000 organizations to small businesses. This wave of transformation is being driven by the ongoing advances in computing building blocks of the last few decades: compute, storage, and networks. At the same time as companies are continuing down the path of digital transformation, there is a new generation of software building blocks that will drive even greater transformation for businesses in the future of which AI is one of the core drivers. AI will help to transform all industries, including transportation, manufacturing, retail, agriculture, and more and create opportunities we have yet to consider. In this post I will talk about how Microsoft is helping businesses transform with AI. This moves beyond creating the AI platform, or using AI to enhance an existing application or service and into how people are starting to think about using AI to change core business processes. If you have been following along, this is the third and final part in a 3-part series on AI. If you want to revisit the other two posts before continuing, Part 1 provides some context on why this time is different for AI. It is followed up by Part 2 whichPart 2 looks at how Microsoft thinks about AI and the tools and services that we make available for you to create your own AI solutions and in turn how we use it to enhance our own products and services. In this blog we’ll dive more into the patterns we see emerging for the use of AI across industries and experiences. As we talk about AI with business and industry leaders, we see 4 patterns emerging on how to apply AI to businesses and business scenarios that provide a useful frame for the conversation.
  1. Virtual Agents - The first pattern is the use of virtual agents to interact with employees, customers, and partners on behalf of a company. These agents can help answer questions, provide support and over time become a proactive representative of your company and your brand. Today Microsoft uses a virtual agent in customer support that will engage in close to 50 million conversations this year.
  2. Ambient Intelligence - The second pattern is anchored on tracking people and objects in a physical space. In many ways this is using AI to both map a physical space and activity to a digital space, and then allowing actions on top of the digital graph. Many people will think about “pickup and go” retail shopping experience as a prime example, but this pattern is also applicable to safety, manufacturing, construction scenarios and more. Think about a warehouse that can detect a person walking in one aisle and a forklift driving in another that are on a collision course, the AI can prevent the pending accident. We also showed this pattern applied to an office/meeting scenario at BUILD.
  3. AI-Assisting Professionals - AI can be used to help almost any professional be more effective. For example, we can help people in finance with forecasting. Large companies manage forecasts by having their front-line sellers predict what they are going to do, then having many layers of reviewers to help judge the forecast. Using historical data and global insights from Bing, LinkedIn, and custom data sources, an AI system can reliably forecast how a subsidiary will do while removing all of the middle layers of judgement and the time consumed doing that. We also see AI starting to assist doctors in areas like genomics and public health. There are great examples in sales, marketing, legal and practically every other profession.
  4. Autonomous Systems - The fourth pattern that we see is for autonomous systems. You might think of self-driving cars when you think about autonomous systems, but it also extends to robotic process automation and network protection. Threats to a network can be hard to identify when they are happening and the lag before responding can result in a lot of damage. Having the network automatically respond as a threat is happening can minimize the risk and free the team to focus on other tasks.
In this post I will cover these 4 patterns along with how Microsoft is helping businesses achieve their goals.

Virtual Agents

At Ignite 2017 we discussed how AI will be used to help large businesses with customer support. Since that time, the Microsoft AI Solution for Customer Service, which is being used by Microsoft, HP, Macy’s, Australian Government Department of Human Services, and others has shown tremendous success in using AI to transform customer engagement. Looking at the initial results from the early adopters we have seen great improvements for their businesses.
  • Microsoft – where possible we are testing AI Solutions within our own business processes before releasing software for others to use (e.g. forecasting and customer care). We built this solution to solve our own problems first and it is trusted to handle one of the largest enterprise support organizations in the world. With the addition of a virtual agent, it has created some impressive results. Over a 6-month period we saw a 2x increase in users successfully being able to help themselves with a 3x decrease in transfers to agents. Users dislike having to repeat themselves as they get transferred between agents, but with this solution the state transfers with the call creating a better experience for customers and our call center employees, at scale with over 100,000 virtual agent sessions per day.
  • HP – one of the early users of this solution has scaled up to handle 70% of support cases with AI and maintains a greater than 85% accurate dialog rate. To achieve this level of accuracy the service was initially trained on over 1 million chat logs and 50 KB of support articles to create a sophisticated dialog.
  • Macy’s – gives an example of how an AI solution for customer support can evolve to become a brand ambassador for the company. The virtual agent is integrated into both the web and mobile web experiences for shopping, so users can use the agents where they are. Macy’s integrated their backend APIs for promotions, alerts, and more into the agent so that the agent could go beyond the corpus of data that it was trained on to pull up account details, shipping updates, and more providing broad and personalized information.
Last year at Ignite when we discussed the Microsoft AI Solution for Customer Service it was framed in the context of helping users get the support they need while also improving the experience for agents who would get to focus on more value-added support cases. Since then there has been an expansion into more personalized experiences as demonstrated by what Macy’s is doing. Looking forward, business agents will be moving towards conversational commerce and blending proactive conversation capability with the current reactive capability. Today online commerce is primarily a self-help experience with the user either having to know exactly what they are looking for or spending a lot of time researching. With a virtual agent there is the opportunity to have a virtual personal shopper by your side who can help you with your online experience. As you search for products the agent can intelligently recommend relevant intents based on what you have been looking at. As you select options that are intelligently designed to guide you through the purchase experience, the information that you need is presented to you in adaptive cards that contains rich yet focused information so that the user can make better decisions without information overload. Since the agent can maintain historical information, a conversation could be resumed later, and transactions could be completed all with the use of the agent. This pattern is by far one of the furthest along and an area we see a lot of activity in today. Companies can choose from a set of tools in the platform today to get started with initial agents before moving up to a full customer care solution.

Ambient Intelligence

Computing is often thought of in the context of devices but with the use of sensors a physical space can be digitized which creates an environment where people, objects and activities can be detected and tracked. When AI is added to the digitized environment you then make it possible to reimagine how a room interacts with the people and objects in it. For example, using facial recognition to know when a person has entered a room is valuable not just for personalization but also for safety. If the fire alarm goes off, knowing who was in the building and has left, or if anyone is still inside can be quite useful. The next level beyond just tracking people is to track people’s interaction with objects around them. This capability makes it possible to create several interesting retail experiences including; grab and go shopping, personalized offers, immediate and on the spot assistance when needed, and more. In manufacturing we can use this type of capability for health and safety scenarios that can include keeping a person safe by flagging when they are going to pick up something that might too heavy, or to make the problem of losing things less burdensome by flagging where a lost item might be with instructions on how to find it and more. There are many interesting scenarios enabled by the ambient pattern that are just starting to be explored. We see real-world examples of this today using mixed reality to keep mission critical systems up and running. Preserving perishable goods is a great example. Cows produce 6 gallons of milk daily, so it is important that the milk is packaged on time to avoid waste. However, when a milk packaging line fails because of a faulty part, it can take several days to get up and running resulting in a lot of spoiled milk. Tetra Pak uses cloud-connected predictive analytics to analyze packaging lines data to predict maintenance issues to reduce down time. This is done by using mixed reality headsets to cut down fixing time by having remote experts guide service engineers.

AI Assisting Professionals

AI presents a great opportunity to augment human ingenuity by providing proactive timely support to people so that they can focus their energy on their most important tasks. This shows up in a variety of ways including digital assistants like Cortana, Alexa, Siri, and Google Home that many of us use today to do tasks for us. In business settings we have seen many early uses of AI helping busy professionals including Bing for Business which will take your companies organizational chart, internal sites, and cloud documents, and then integrate them into your search experience so that you have the public and private data that you need. AI helps professionals in other ways more specific to different industries. Attorneys who need to assess the risk of an event will spend a lot of time going through contracts to find the impact of any exposure to a negative event. Machine reading techniques can be used to understand the details of all their contracts and then surface the problems. Journalists can use AI to serve as a virtual editor that will look beyond rules-based spellchecking to make writing suggestions. Finance professionals can also use AI to do sales forecasts to make better forecasts or sellers can use AI to target sales prospects and how to close a deal. Marketers can use AI to predict new trends in customer interest before investing a lot of money to experiment. We are working on a variety of areas where we can apply AI to assist professionals and one of my favorite examples is the work we are doing to help medical professionals. Microsoft Research is known for world class advances in computer science, but they also focus on medical, health, and genomics. Their approach is to apply novel computational tools and analytical techniques to make healthcare more impactful and to assist patients. One application of AI to healthcare is Project InnerEye. Today expert medical practitioners spend a lot of time analyzing 3D images to identify tumors versus other healthy tissue. Using years of Microsoft AI research in computer vision including deep decision forests which are used in the Kinect and HoloLens, and machine learning, and then applying it to more medical images than the average medical practitioner could analyze on their own, Project InnerEye can assist in identifying the presence of a tumor. To ensure that the medical practitioner is in charge, the results can be adjusted by the experts until everyone is comfortable with the result. Since we are a platform company we are also making the technology available to third-party medical software companies so that medical practitioners can use the tools they are most comfortable with today.

Autonomous Systems

Most people when they think of autonomous systems, visions of self-driving cars come up but there are so many other ways to apply this technology as well. Networks when they are under attack can take a long time for someone to notice and even longer to determine the root cause and then address it. Machine learning does a great job of detecting patterns in a vast amount of data, so it can be used to quickly identify attacks in real-time and then AI can be used to find the most effective approach for addressing the problem. The Network Admin will be notified and can control the process if needed but speeding up this process will limit the amount of damage that an attacker could cause. At the RSA 2018 Conference, Mark Russinovich shared how Microsoft’s investment in AI has created new security capabilities for protecting all of our customers. For security, autonomous systems can help create secure products by finding security flaws during development. Microsoft Security Risk Detection is our offering for customers to use AI to find security flaws. DocuSign is a company that enables people to sign documents from any device which means that they take security very seriously. Like most software developers, DocuSign uses a variety of software components in their products including ones from third parties and the security of their product is only as strong as all of the components put together. Using Microsoft Security Risk Detection, DocuSign can automatically check the security of all their components across multiple virtual machines with no extra work.
“We were able to automatically run millions of test cases across multiple virtual machines, entirely automated, with no extra work …. We’re always looking for services that add value, and that scale of automation is a great added value.” — John Heasman, Senior Director, Software Security, DocuSign
It is an exciting time in the industry as artificial intelligence is not just a topic of conversation but becoming the starting point for thinking about what will become the next wave(s) of digital and/or AI based transformation. The examples presented here across the four patterns are just a few examples, but we see the patterns as an interesting point to start the conversation. Three weeks I ago I started this series discussing why this time is different for AI compared to other times of AI excitement in the New Generation of Software Building Blocks post. Last week we focused on Microsoft’s approach to AI and how we are making it available to others across the platform, our products, and solutions. Today we presented four patterns for the use of AI in businesses and industries that we hope provide a useful framing around the topic. As you think through your own use cases think about how these patterns can apply to your business. Thank you for taking the time to read the series and if there are any other topics that we should explore please let us know.   Cheers, Guggs]]> Microsoft’s Approach to AI https://docs.microsoft.com/archive/blogs/stevengu/microsofts-approach-to-ai Fri, 18 May 2018 08:00:20 GMT https://blogs.msdn.microsoft.com/stevengu/?p=1675 post I wrote about the new generation of technology building blocks that will shape the future of digital experiences with a specific focus on Artificial Intelligence, and what is driving the rapid expansion of this capability. In this post I will build on that conversation to discuss how Microsoft currently approaches AI from a business perspective.  Before we go into the details, be sure to revisit the first post in this series for context on why AI is taking off now. At Microsoft there are teams working on AI projects around the company which generally falls into 3 core categories of platform and product investment. AI Platform & Services – We are focused on building a new set of AI services and tools to make AI accessible to every organization.  At Microsoft we are platform builders by trade, that is we build the infrastructure that others use to build their products and services on.  In the case of Artificial Intelligence, we are creating the infrastructure, services, and tools that allow developers and data scientists to infuse AI into their applications and services, as well as build new and unique solutions that are AI based. Infusing AI – The second area of focus is looking at how we can make our core products better by infusing them with AI.  A generation ago, industry embraced the Internet, weaving it into virtually every product and service. The same is now happening with AI.  In many cases you might not even know it’s there – but it’s helping individuals and organizations connect and make more informed decisions. Business Solutions – The third area is looking at how we start with deep AI capabilities to build a new generation of AI-based business solutions.  We are looking at this first in terms of how we can use AI to help make people and organizations more productive, starting with some of our own processes that are used to run Microsoft. These three areas represent an exciting period for bringing AI to life in a very pragmatic way.  These efforts build on top of, and will continue to build on, the core set of research that is going on in the field of Artificial Intelligence by our Research group and others in the industry.

AI Platform & Services

It’s invigorating to watch the birth of a new platform or platform layer, which is what we are witnessing in AI right now. There are a set of tools and services that started out as pure research (see last week’s post) that are being normalized for use by the broad developer base and data scientists.  This is an important step in the democratization of AI, as today it’s still difficult for organizations to find and hire people who are versed in the science of AI.  Previous AI systems were primarily sophisticated rules-based engines, but with the growth and success of deep neural networks, and the ability to rapidly test and deploy AI algorithms, the baseline background for AI practitioners has changed.  As a result, the state of the art for AI is quickly advancing, making it harder to find people who can be subject matter experts across the growing set of AI capabilities.  Searching for AI research papers released in 2018 already shows over 800 published papers.  In that light, the need for a platform layer that brings AI in a consistent fashion to traditional development teams is critical for getting started with AI for their use cases. The Microsoft AI platform consists of infrastructure to provide AI at scale, services that provide core AI capabilities through a common set of APIs, and tools for practitioners who want to be hands on in the creation of their own custom AI models. Customers can now test and train at scale on CPUs and GPUs, and soon, will be able to tap the power of FPGAs – software programmable chips such as those powering Microsoft’s Project Brainwave and Jabil’s factory of the future as featured at our BUILD developer conference just last week. Already FPGAs offer a 132x speed enhancement over CPUs and being software programmable, FPGAs offer a balance of performance and future-proofing.  As new innovations happen in AI methods, they can more quickly be deployed via software to the FPGA’s in contrast to building custom silicon such as the ASICs powering TPUs. As a result, time to market and deployment of these new innovations can be significantly reduced. FPGAs are now powering Bing search, and we’re seeing 5x lower hardware latency than TPUs for real-time AI and we are working to deploy FPGAs for customers as a part of the Intelligent Cloud, and Intelligent Edge devices.
  1. Pre-Trained AI Services - For most developers and organizations starting their AI journey, the core of this new platform is a set of Azure-powered Cognitive Services that help developers bring human cognition (the ability to speak, hear, translate, see, reason, etc.) into their applications. Taking the research breakthroughs and turning them into a consistent set of developer services takes some time, but we are seeing the fruit of that work with a growing set of cohesive services, with consistency in APIs, sample code, documentation, language support and more.  These services give developers pre-built AI models, exposed as services for vision, speech, language, conversation, search, and knowledge.  Many developers are using these services for creating great user experiences and solving business problems rather than spending the time to train their own deep neural networks.  These core services work for a wide variety of developer needs and are also customizable for additional specifics an experience might require.  Azure-powered services such as Custom Vision make it possible to upload labeled images, train with a few dozen images, then evaluate and learn to make your classifier more specific to a given task or process, and then even deploy to mobile devices such as Windows, iOS, and Android.  The same can be done with custom language domains…say for lawyers or doctors.   Cognitive Services provide the best of general purpose capability and customizability with speed to solution.
  2. Conversational AI - Hand in hand with Cognitive Services, we are seeing a transition to a more conversational user interaction as people start to infuse AI into their solutions. Providing the ability for developers to add naturally fluid conversations as part of an overall user experience, across canvasses and state, is the goal of the Azure Bot Services. With the Azure Bot Service and LUIS for language understanding, you can create, deploy and manage a bot that interacts with your users on the channels they already are in your apps, Facebook Messenger, Slack, Skype, Microsoft Teams, Websites, Cortana, and more. When trying to create a more natural conversational experience the Azure Bot Service can be combined with the Cognitive Services such as translation or custom vision to expand your bot’s capabilities over time.
  3. Tools and Open Formats - While these services and tools can be accessed independently they are also being integrated into some of our core tools such as the Azure Portal and Visual Studio. Developers and data scientists that want to work a layer deeper on the fundamental science, they can through support for key frameworks and tools within our platform.  Since AI is still rapidly evolving some developers will be interested in researching new techniques or doing something very specific; like creating a generative adversarial network (GAN) or a very deep neural network. In those cases, people will want to use tools like Cognitive Toolkit, TensorFlow, Caffe, MXNet, Chainer, and more to train and deploy their own custom models.  Azure is committed to support a broad range of these AI specific tool chains.  To help in this space Microsoft is the co-founder of the Open Neural Network Exchange (ONNX) along with Facebook, which is an open format for exchanging deep learning models between different toolchains. Since announcing ONNX, other tech companies have joined including Amazon, AMD, ARM, Baidu, Huawei, IBM, Intel, Nvidia, Qualcomm, and many more.
  4. Extending to the Intelligent Edge - There is also a lot of great learning around the importance of edge support for AI as we delve further into world solutions and use cases. In this area Microsoft is focused on enabling AI at the intelligent edge through a variety of tools and offerings. There were several announcements around BUILD that highlight the effort to create an intelligent edge that supports real world AI scenarios.  These include:

Infusing AI into our Products and Services

Our second area of focus as a company is how can we make our products better by infusing them with AI.  The goal of course is to create a better customer experience, and if done well, you don’t know that the products are infused with AI, they just work better.  This work can be done either behind the scenes or directly in the experience.  Behind the scenes, we use AI to run our networks more effectively; like Skype or Azure, or to protect people’s data and content in the cloud or on devices.  Within our products we also strive to create unique experiences that would not be possible without AI. For example, the resume assistant in Word helps the author create a resume that identifies the top skills that are competitive with other people in the same field.  PowerPoint designer enables anyone creating a slide with images to make them visually interesting regardless of their artistic ability.  Excel Insights helps visualize patterns in your data, and the new Acronyms feature in Word that helps master company lingo with Machine Learning.   Using Natural Language Processing and Language Understanding to allow real time conversations between people in two different languages over Skype is bridging communities around the world.  These same tools can be used to enable closed captioning of a speech in PowerPoint, and even do the captioning in multiple languages at the same time.  In Windows Hello, the ability to securely login to your PC just by looking at it is based on AI.  There is a growing number of use cases, and if AI is properly infused into a product it should just work better.

Solutions

The third area we focus on is how can companies use AI to improve their overall business.  Whether it’s how products are built, customers and partners are helped, or employees are cared for, AI can support a broad set of standard use cases.  In this area Microsoft has been using AI internally on a variety of processes that range from forecasting to marketing to customer care and more.  As we spend time in this area and worked with a broad range of customers on their specific solutions, we’ve started to see a set of patterns emerge that reflect a lot of the focus of AI within organizations. I find these patterns of virtual agents, ambient intelligence, AI-assisted professional, and autonomous systems to be helpful in having conversations with customers/partners around the potential opportunities to get started with AI for their business. In my next blog I will go deeper into these four patterns. When it comes to AI, Microsoft is vested in creating a platform to allow developers and companies to infuse AI into their services and products, we are also using AI to create better products for our customers and finally we are using AI to help better run our company and help other companies better run theirs.  This is Microsoft’s approach for making AI available for others.  Join me next week for the third and final part of this series where I will delve a little deeper into how we are seeing customers adopt AI. If you missed the first part of the series, you can read it here.   Cheers, Guggs]]>